I am working on perceptual image hashing, i convert image to grey scale in c# windowform now i want to compute hash value of image. i don't want to use SHA,MD5 etc as these are VERY sensitive to input changes.The algorithm should be such that it is not affected by small changes in the input image. This is the entire problem.We need to perform OCR,i.e. convert image into text. Images can change for a variety of reasons, such as page condition, camera focus, lighting etc. As humans, we can still read text in the images even if it is affected by any or all of these problems. We need to devise an algorithm that does the same thing, i.e. if the input image is changed slightly, our hash algorithm still generates the same or very close hash for it.
Asked
Active
Viewed 367 times
0
-
Not a duplicate but I've done something similar [here](https://stackoverflow.com/questions/35151067/algorithm-to-compare-two-images-in-c-sharp/35153895#35153895) – fubo Mar 27 '18 at 06:51
-
A very close hash for a similar image? Looks like you aren't really looking for a *hash* but for some kind of *fingerprint* that you can take to compare the similarity of images. Out of the box I would think of this fingerprint as a coordinate in a x-dimensional configuration space where x is the number of properties that you want to take into account (e.g. brightness, dominant color etc). – Rob Mar 27 '18 at 08:53